; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -mtriple=x86_64-unknown-linux -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,DEFAULT,SSE
; RUN: opt < %s -mtriple=x86_64-unknown-linux -mcpu=corei7-avx -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,DEFAULT,AVX
; RUN: opt < %s -mtriple=x86_64-unknown-linux -mcpu=core-avx2 -slp-vectorizer -S | FileCheck %s --check-prefixes=CHECK,DEFAULT,AVX2
; RUN: opt < %s -mtriple=x86_64-unknown-linux -mcpu=skx -slp-vectorizer -S -slp-threshold=-100 | FileCheck %s --check-prefixes=CHECK,THRESH

@arr = local_unnamed_addr global [32 x i32] zeroinitializer, align 16
@arr1 = local_unnamed_addr global [32 x float] zeroinitializer, align 16
@arrp = local_unnamed_addr global [32 x i32*] zeroinitializer, align 16
@var = global i32 zeroinitializer, align 8

declare i32 @llvm.smax.i32(i32, i32)
declare i16 @llvm.smin.i16(i16, i16)
declare i64 @llvm.umax.i64(i64, i64)
declare i8 @llvm.umin.i8(i8, i8)
declare i32 @llvm.smin.i32(i32, i32)
declare i32 @llvm.umin.i32(i32, i32)

define i32 @maxi8(i32) {
; SSE-LABEL: @maxi8(
; SSE-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
; SSE-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
; SSE-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
; SSE-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
; SSE-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
; SSE-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
; SSE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 [[TMP5]], i32 [[TMP6]]
; SSE-NEXT:    [[TMP9:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
; SSE-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP8]], i32 [[TMP9]]
; SSE-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
; SSE-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
; SSE-NEXT:    [[TMP15:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
; SSE-NEXT:    [[TMP16:%.*]] = icmp sgt i32 [[TMP14]], [[TMP15]]
; SSE-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP14]], i32 [[TMP15]]
; SSE-NEXT:    [[TMP18:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; SSE-NEXT:    [[TMP19:%.*]] = icmp sgt i32 [[TMP17]], [[TMP18]]
; SSE-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP17]], i32 [[TMP18]]
; SSE-NEXT:    [[TMP21:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; SSE-NEXT:    [[TMP22:%.*]] = icmp sgt i32 [[TMP20]], [[TMP21]]
; SSE-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP20]], i32 [[TMP21]]
; SSE-NEXT:    ret i32 [[TMP23]]
;
; AVX-LABEL: @maxi8(
; AVX-NEXT:    [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([32 x i32]* @arr to <8 x i32>*), align 16
; AVX-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> [[TMP2]])
; AVX-NEXT:    ret i32 [[TMP3]]
;
; AVX2-LABEL: @maxi8(
; AVX2-NEXT:    [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([32 x i32]* @arr to <8 x i32>*), align 16
; AVX2-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> [[TMP2]])
; AVX2-NEXT:    ret i32 [[TMP3]]
;
; THRESH-LABEL: @maxi8(
; THRESH-NEXT:    [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([32 x i32]* @arr to <8 x i32>*), align 16
; THRESH-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> [[TMP2]])
; THRESH-NEXT:    ret i32 [[TMP3]]
;
  %2 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
  %3 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
  %4 = icmp sgt i32 %2, %3
  %5 = select i1 %4, i32 %2, i32 %3
  %6 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
  %7 = icmp sgt i32 %5, %6
  %8 = select i1 %7, i32 %5, i32 %6
  %9 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
  %10 = icmp sgt i32 %8, %9
  %11 = select i1 %10, i32 %8, i32 %9
  %12 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
  %13 = icmp sgt i32 %11, %12
  %14 = select i1 %13, i32 %11, i32 %12
  %15 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
  %16 = icmp sgt i32 %14, %15
  %17 = select i1 %16, i32 %14, i32 %15
  %18 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
  %19 = icmp sgt i32 %17, %18
  %20 = select i1 %19, i32 %17, i32 %18
  %21 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
  %22 = icmp sgt i32 %20, %21
  %23 = select i1 %22, i32 %20, i32 %21
  ret i32 %23
}

define i32 @maxi8_store_in(i32) {
; SSE-LABEL: @maxi8_store_in(
; SSE-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
; SSE-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
; SSE-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
; SSE-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
; SSE-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
; SSE-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
; SSE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 [[TMP5]], i32 [[TMP6]]
; SSE-NEXT:    [[TMP9:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
; SSE-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP8]], i32 [[TMP9]]
; SSE-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
; SSE-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
; SSE-NEXT:    store i32 0, i32* @var, align 8
; SSE-NEXT:    [[TMP15:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
; SSE-NEXT:    [[TMP16:%.*]] = icmp sgt i32 [[TMP14]], [[TMP15]]
; SSE-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP14]], i32 [[TMP15]]
; SSE-NEXT:    [[TMP18:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; SSE-NEXT:    [[TMP19:%.*]] = icmp sgt i32 [[TMP17]], [[TMP18]]
; SSE-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP17]], i32 [[TMP18]]
; SSE-NEXT:    [[TMP21:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; SSE-NEXT:    [[TMP22:%.*]] = icmp sgt i32 [[TMP20]], [[TMP21]]
; SSE-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP20]], i32 [[TMP21]]
; SSE-NEXT:    ret i32 [[TMP23]]
;
; AVX-LABEL: @maxi8_store_in(
; AVX-NEXT:    [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([32 x i32]* @arr to <8 x i32>*), align 16
; AVX-NEXT:    store i32 0, i32* @var, align 8
; AVX-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> [[TMP2]])
; AVX-NEXT:    ret i32 [[TMP3]]
;
; AVX2-LABEL: @maxi8_store_in(
; AVX2-NEXT:    [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([32 x i32]* @arr to <8 x i32>*), align 16
; AVX2-NEXT:    store i32 0, i32* @var, align 8
; AVX2-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> [[TMP2]])
; AVX2-NEXT:    ret i32 [[TMP3]]
;
; THRESH-LABEL: @maxi8_store_in(
; THRESH-NEXT:    [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([32 x i32]* @arr to <8 x i32>*), align 16
; THRESH-NEXT:    store i32 0, i32* @var, align 8
; THRESH-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> [[TMP2]])
; THRESH-NEXT:    ret i32 [[TMP3]]
;
  %2 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
  %3 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
  %4 = icmp sgt i32 %2, %3
  %5 = select i1 %4, i32 %2, i32 %3
  %6 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
  %7 = icmp sgt i32 %5, %6
  %8 = select i1 %7, i32 %5, i32 %6
  %9 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
  %10 = icmp sgt i32 %8, %9
  %11 = select i1 %10, i32 %8, i32 %9
  %12 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
  %13 = icmp sgt i32 %11, %12
  %14 = select i1 %13, i32 %11, i32 %12
  store i32 0, i32* @var, align 8
  %15 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
  %16 = icmp sgt i32 %14, %15
  %17 = select i1 %16, i32 %14, i32 %15
  %18 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
  %19 = icmp sgt i32 %17, %18
  %20 = select i1 %19, i32 %17, i32 %18
  %21 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
  %22 = icmp sgt i32 %20, %21
  %23 = select i1 %22, i32 %20, i32 %21
  ret i32 %23
}

define i32 @maxi16(i32) {
; CHECK-LABEL: @maxi16(
; CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i32>, <16 x i32>* bitcast ([32 x i32]* @arr to <16 x i32>*), align 16
; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v16i32(<16 x i32> [[TMP2]])
; CHECK-NEXT:    ret i32 [[TMP3]]
;
  %2 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
  %3 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
  %4 = icmp sgt i32 %2, %3
  %5 = select i1 %4, i32 %2, i32 %3
  %6 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
  %7 = icmp sgt i32 %5, %6
  %8 = select i1 %7, i32 %5, i32 %6
  %9 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
  %10 = icmp sgt i32 %8, %9
  %11 = select i1 %10, i32 %8, i32 %9
  %12 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
  %13 = icmp sgt i32 %11, %12
  %14 = select i1 %13, i32 %11, i32 %12
  %15 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
  %16 = icmp sgt i32 %14, %15
  %17 = select i1 %16, i32 %14, i32 %15
  %18 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
  %19 = icmp sgt i32 %17, %18
  %20 = select i1 %19, i32 %17, i32 %18
  %21 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
  %22 = icmp sgt i32 %20, %21
  %23 = select i1 %22, i32 %20, i32 %21
  %24 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 8), align 16
  %25 = icmp sgt i32 %23, %24
  %26 = select i1 %25, i32 %23, i32 %24
  %27 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 9), align 4
  %28 = icmp sgt i32 %26, %27
  %29 = select i1 %28, i32 %26, i32 %27
  %30 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 10), align 8
  %31 = icmp sgt i32 %29, %30
  %32 = select i1 %31, i32 %29, i32 %30
  %33 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 11), align 4
  %34 = icmp sgt i32 %32, %33
  %35 = select i1 %34, i32 %32, i32 %33
  %36 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 12), align 16
  %37 = icmp sgt i32 %35, %36
  %38 = select i1 %37, i32 %35, i32 %36
  %39 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 13), align 4
  %40 = icmp sgt i32 %38, %39
  %41 = select i1 %40, i32 %38, i32 %39
  %42 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 14), align 8
  %43 = icmp sgt i32 %41, %42
  %44 = select i1 %43, i32 %41, i32 %42
  %45 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 15), align 4
  %46 = icmp sgt i32 %44, %45
  %47 = select i1 %46, i32 %44, i32 %45
  ret i32 %47
}

define i32 @maxi32(i32) {
; CHECK-LABEL: @maxi32(
; CHECK-NEXT:    [[TMP2:%.*]] = load <32 x i32>, <32 x i32>* bitcast ([32 x i32]* @arr to <32 x i32>*), align 16
; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v32i32(<32 x i32> [[TMP2]])
; CHECK-NEXT:    ret i32 [[TMP3]]
;
  %2 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
  %3 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
  %4 = icmp sgt i32 %2, %3
  %5 = select i1 %4, i32 %2, i32 %3
  %6 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
  %7 = icmp sgt i32 %5, %6
  %8 = select i1 %7, i32 %5, i32 %6
  %9 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
  %10 = icmp sgt i32 %8, %9
  %11 = select i1 %10, i32 %8, i32 %9
  %12 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
  %13 = icmp sgt i32 %11, %12
  %14 = select i1 %13, i32 %11, i32 %12
  %15 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
  %16 = icmp sgt i32 %14, %15
  %17 = select i1 %16, i32 %14, i32 %15
  %18 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
  %19 = icmp sgt i32 %17, %18
  %20 = select i1 %19, i32 %17, i32 %18
  %21 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
  %22 = icmp sgt i32 %20, %21
  %23 = select i1 %22, i32 %20, i32 %21
  %24 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 8), align 16
  %25 = icmp sgt i32 %23, %24
  %26 = select i1 %25, i32 %23, i32 %24
  %27 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 9), align 4
  %28 = icmp sgt i32 %26, %27
  %29 = select i1 %28, i32 %26, i32 %27
  %30 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 10), align 8
  %31 = icmp sgt i32 %29, %30
  %32 = select i1 %31, i32 %29, i32 %30
  %33 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 11), align 4
  %34 = icmp sgt i32 %32, %33
  %35 = select i1 %34, i32 %32, i32 %33
  %36 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 12), align 16
  %37 = icmp sgt i32 %35, %36
  %38 = select i1 %37, i32 %35, i32 %36
  %39 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 13), align 4
  %40 = icmp sgt i32 %38, %39
  %41 = select i1 %40, i32 %38, i32 %39
  %42 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 14), align 8
  %43 = icmp sgt i32 %41, %42
  %44 = select i1 %43, i32 %41, i32 %42
  %45 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 15), align 4
  %46 = icmp sgt i32 %44, %45
  %47 = select i1 %46, i32 %44, i32 %45
  %48 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 16), align 16
  %49 = icmp sgt i32 %47, %48
  %50 = select i1 %49, i32 %47, i32 %48
  %51 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 17), align 4
  %52 = icmp sgt i32 %50, %51
  %53 = select i1 %52, i32 %50, i32 %51
  %54 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 18), align 8
  %55 = icmp sgt i32 %53, %54
  %56 = select i1 %55, i32 %53, i32 %54
  %57 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 19), align 4
  %58 = icmp sgt i32 %56, %57
  %59 = select i1 %58, i32 %56, i32 %57
  %60 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 20), align 16
  %61 = icmp sgt i32 %59, %60
  %62 = select i1 %61, i32 %59, i32 %60
  %63 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 21), align 4
  %64 = icmp sgt i32 %62, %63
  %65 = select i1 %64, i32 %62, i32 %63
  %66 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 22), align 8
  %67 = icmp sgt i32 %65, %66
  %68 = select i1 %67, i32 %65, i32 %66
  %69 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 23), align 4
  %70 = icmp sgt i32 %68, %69
  %71 = select i1 %70, i32 %68, i32 %69
  %72 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 24), align 16
  %73 = icmp sgt i32 %71, %72
  %74 = select i1 %73, i32 %71, i32 %72
  %75 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 25), align 4
  %76 = icmp sgt i32 %74, %75
  %77 = select i1 %76, i32 %74, i32 %75
  %78 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 26), align 8
  %79 = icmp sgt i32 %77, %78
  %80 = select i1 %79, i32 %77, i32 %78
  %81 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 27), align 4
  %82 = icmp sgt i32 %80, %81
  %83 = select i1 %82, i32 %80, i32 %81
  %84 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 28), align 16
  %85 = icmp sgt i32 %83, %84
  %86 = select i1 %85, i32 %83, i32 %84
  %87 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 29), align 4
  %88 = icmp sgt i32 %86, %87
  %89 = select i1 %88, i32 %86, i32 %87
  %90 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 30), align 8
  %91 = icmp sgt i32 %89, %90
  %92 = select i1 %91, i32 %89, i32 %90
  %93 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 31), align 4
  %94 = icmp sgt i32 %92, %93
  %95 = select i1 %94, i32 %92, i32 %93
  ret i32 %95
}

; Note: legacy test - InstCombine creates maxnum intrinsics for fcmp+select with fastmath on the select.

define float @maxf8(float) {
; DEFAULT-LABEL: @maxf8(
; DEFAULT-NEXT:    [[TMP2:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 0), align 16
; DEFAULT-NEXT:    [[TMP3:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 1), align 4
; DEFAULT-NEXT:    [[TMP4:%.*]] = fcmp fast ogt float [[TMP2]], [[TMP3]]
; DEFAULT-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], float [[TMP2]], float [[TMP3]]
; DEFAULT-NEXT:    [[TMP6:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 2), align 8
; DEFAULT-NEXT:    [[TMP7:%.*]] = fcmp fast ogt float [[TMP5]], [[TMP6]]
; DEFAULT-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], float [[TMP5]], float [[TMP6]]
; DEFAULT-NEXT:    [[TMP9:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 3), align 4
; DEFAULT-NEXT:    [[TMP10:%.*]] = fcmp fast ogt float [[TMP8]], [[TMP9]]
; DEFAULT-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], float [[TMP8]], float [[TMP9]]
; DEFAULT-NEXT:    [[TMP12:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 4), align 16
; DEFAULT-NEXT:    [[TMP13:%.*]] = fcmp fast ogt float [[TMP11]], [[TMP12]]
; DEFAULT-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], float [[TMP11]], float [[TMP12]]
; DEFAULT-NEXT:    [[TMP15:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 5), align 4
; DEFAULT-NEXT:    [[TMP16:%.*]] = fcmp fast ogt float [[TMP14]], [[TMP15]]
; DEFAULT-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], float [[TMP14]], float [[TMP15]]
; DEFAULT-NEXT:    [[TMP18:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 6), align 8
; DEFAULT-NEXT:    [[TMP19:%.*]] = fcmp fast ogt float [[TMP17]], [[TMP18]]
; DEFAULT-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], float [[TMP17]], float [[TMP18]]
; DEFAULT-NEXT:    [[TMP21:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 7), align 4
; DEFAULT-NEXT:    [[TMP22:%.*]] = fcmp fast ogt float [[TMP20]], [[TMP21]]
; DEFAULT-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], float [[TMP20]], float [[TMP21]]
; DEFAULT-NEXT:    ret float [[TMP23]]
;
; THRESH-LABEL: @maxf8(
; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x float>, <2 x float>* bitcast ([32 x float]* @arr1 to <2 x float>*), align 16
; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x float> [[TMP2]], i32 1
; THRESH-NEXT:    [[TMP5:%.*]] = fcmp fast ogt float [[TMP3]], [[TMP4]]
; THRESH-NEXT:    [[TMP6:%.*]] = select i1 [[TMP5]], float [[TMP3]], float [[TMP4]]
; THRESH-NEXT:    [[TMP7:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 2), align 8
; THRESH-NEXT:    [[TMP8:%.*]] = fcmp fast ogt float [[TMP6]], [[TMP7]]
; THRESH-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], float [[TMP6]], float [[TMP7]]
; THRESH-NEXT:    [[TMP10:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 3), align 4
; THRESH-NEXT:    [[TMP11:%.*]] = fcmp fast ogt float [[TMP9]], [[TMP10]]
; THRESH-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], float [[TMP9]], float [[TMP10]]
; THRESH-NEXT:    [[TMP13:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 4), align 16
; THRESH-NEXT:    [[TMP14:%.*]] = fcmp fast ogt float [[TMP12]], [[TMP13]]
; THRESH-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], float [[TMP12]], float [[TMP13]]
; THRESH-NEXT:    [[TMP16:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 5), align 4
; THRESH-NEXT:    [[TMP17:%.*]] = fcmp fast ogt float [[TMP15]], [[TMP16]]
; THRESH-NEXT:    [[TMP18:%.*]] = select i1 [[TMP17]], float [[TMP15]], float [[TMP16]]
; THRESH-NEXT:    [[TMP19:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 6), align 8
; THRESH-NEXT:    [[TMP20:%.*]] = fcmp fast ogt float [[TMP18]], [[TMP19]]
; THRESH-NEXT:    [[TMP21:%.*]] = select i1 [[TMP20]], float [[TMP18]], float [[TMP19]]
; THRESH-NEXT:    [[TMP22:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 7), align 4
; THRESH-NEXT:    [[TMP23:%.*]] = fcmp fast ogt float [[TMP21]], [[TMP22]]
; THRESH-NEXT:    [[TMP24:%.*]] = select i1 [[TMP23]], float [[TMP21]], float [[TMP22]]
; THRESH-NEXT:    ret float [[TMP24]]
;
  %2 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 0), align 16
  %3 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 1), align 4
  %4 = fcmp fast ogt float %2, %3
  %5 = select i1 %4, float %2, float %3
  %6 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 2), align 8
  %7 = fcmp fast ogt float %5, %6
  %8 = select i1 %7, float %5, float %6
  %9 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 3), align 4
  %10 = fcmp fast ogt float %8, %9
  %11 = select i1 %10, float %8, float %9
  %12 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 4), align 16
  %13 = fcmp fast ogt float %11, %12
  %14 = select i1 %13, float %11, float %12
  %15 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 5), align 4
  %16 = fcmp fast ogt float %14, %15
  %17 = select i1 %16, float %14, float %15
  %18 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 6), align 8
  %19 = fcmp fast ogt float %17, %18
  %20 = select i1 %19, float %17, float %18
  %21 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 7), align 4
  %22 = fcmp fast ogt float %20, %21
  %23 = select i1 %22, float %20, float %21
  ret float %23
}

; Note: legacy test - maxnum intrinsics match what InstCombine creates for fcmp+select with fastmath on the select.

define float @maxf16(float) {
; DEFAULT-LABEL: @maxf16(
; DEFAULT-NEXT:    [[TMP2:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 0), align 16
; DEFAULT-NEXT:    [[TMP3:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 1), align 4
; DEFAULT-NEXT:    [[TMP4:%.*]] = fcmp fast ogt float [[TMP2]], [[TMP3]]
; DEFAULT-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], float [[TMP2]], float [[TMP3]]
; DEFAULT-NEXT:    [[TMP6:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 2), align 8
; DEFAULT-NEXT:    [[TMP7:%.*]] = fcmp fast ogt float [[TMP5]], [[TMP6]]
; DEFAULT-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], float [[TMP5]], float [[TMP6]]
; DEFAULT-NEXT:    [[TMP9:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 3), align 4
; DEFAULT-NEXT:    [[TMP10:%.*]] = fcmp fast ogt float [[TMP8]], [[TMP9]]
; DEFAULT-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], float [[TMP8]], float [[TMP9]]
; DEFAULT-NEXT:    [[TMP12:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 4), align 16
; DEFAULT-NEXT:    [[TMP13:%.*]] = fcmp fast ogt float [[TMP11]], [[TMP12]]
; DEFAULT-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], float [[TMP11]], float [[TMP12]]
; DEFAULT-NEXT:    [[TMP15:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 5), align 4
; DEFAULT-NEXT:    [[TMP16:%.*]] = fcmp fast ogt float [[TMP14]], [[TMP15]]
; DEFAULT-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], float [[TMP14]], float [[TMP15]]
; DEFAULT-NEXT:    [[TMP18:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 6), align 8
; DEFAULT-NEXT:    [[TMP19:%.*]] = fcmp fast ogt float [[TMP17]], [[TMP18]]
; DEFAULT-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], float [[TMP17]], float [[TMP18]]
; DEFAULT-NEXT:    [[TMP21:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 7), align 4
; DEFAULT-NEXT:    [[TMP22:%.*]] = fcmp fast ogt float [[TMP20]], [[TMP21]]
; DEFAULT-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], float [[TMP20]], float [[TMP21]]
; DEFAULT-NEXT:    [[TMP24:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 8), align 16
; DEFAULT-NEXT:    [[TMP25:%.*]] = fcmp fast ogt float [[TMP23]], [[TMP24]]
; DEFAULT-NEXT:    [[TMP26:%.*]] = select i1 [[TMP25]], float [[TMP23]], float [[TMP24]]
; DEFAULT-NEXT:    [[TMP27:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 9), align 4
; DEFAULT-NEXT:    [[TMP28:%.*]] = fcmp fast ogt float [[TMP26]], [[TMP27]]
; DEFAULT-NEXT:    [[TMP29:%.*]] = select i1 [[TMP28]], float [[TMP26]], float [[TMP27]]
; DEFAULT-NEXT:    [[TMP30:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 10), align 8
; DEFAULT-NEXT:    [[TMP31:%.*]] = fcmp fast ogt float [[TMP29]], [[TMP30]]
; DEFAULT-NEXT:    [[TMP32:%.*]] = select i1 [[TMP31]], float [[TMP29]], float [[TMP30]]
; DEFAULT-NEXT:    [[TMP33:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 11), align 4
; DEFAULT-NEXT:    [[TMP34:%.*]] = fcmp fast ogt float [[TMP32]], [[TMP33]]
; DEFAULT-NEXT:    [[TMP35:%.*]] = select i1 [[TMP34]], float [[TMP32]], float [[TMP33]]
; DEFAULT-NEXT:    [[TMP36:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 12), align 16
; DEFAULT-NEXT:    [[TMP37:%.*]] = fcmp fast ogt float [[TMP35]], [[TMP36]]
; DEFAULT-NEXT:    [[TMP38:%.*]] = select i1 [[TMP37]], float [[TMP35]], float [[TMP36]]
; DEFAULT-NEXT:    [[TMP39:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 13), align 4
; DEFAULT-NEXT:    [[TMP40:%.*]] = fcmp fast ogt float [[TMP38]], [[TMP39]]
; DEFAULT-NEXT:    [[TMP41:%.*]] = select i1 [[TMP40]], float [[TMP38]], float [[TMP39]]
; DEFAULT-NEXT:    [[TMP42:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 14), align 8
; DEFAULT-NEXT:    [[TMP43:%.*]] = fcmp fast ogt float [[TMP41]], [[TMP42]]
; DEFAULT-NEXT:    [[TMP44:%.*]] = select i1 [[TMP43]], float [[TMP41]], float [[TMP42]]
; DEFAULT-NEXT:    [[TMP45:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 15), align 4
; DEFAULT-NEXT:    [[TMP46:%.*]] = fcmp fast ogt float [[TMP44]], [[TMP45]]
; DEFAULT-NEXT:    [[TMP47:%.*]] = select i1 [[TMP46]], float [[TMP44]], float [[TMP45]]
; DEFAULT-NEXT:    ret float [[TMP47]]
;
; THRESH-LABEL: @maxf16(
; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x float>, <2 x float>* bitcast ([32 x float]* @arr1 to <2 x float>*), align 16
; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x float> [[TMP2]], i32 1
; THRESH-NEXT:    [[TMP5:%.*]] = fcmp fast ogt float [[TMP3]], [[TMP4]]
; THRESH-NEXT:    [[TMP6:%.*]] = select i1 [[TMP5]], float [[TMP3]], float [[TMP4]]
; THRESH-NEXT:    [[TMP7:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 2), align 8
; THRESH-NEXT:    [[TMP8:%.*]] = fcmp fast ogt float [[TMP6]], [[TMP7]]
; THRESH-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], float [[TMP6]], float [[TMP7]]
; THRESH-NEXT:    [[TMP10:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 3), align 4
; THRESH-NEXT:    [[TMP11:%.*]] = fcmp fast ogt float [[TMP9]], [[TMP10]]
; THRESH-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], float [[TMP9]], float [[TMP10]]
; THRESH-NEXT:    [[TMP13:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 4), align 16
; THRESH-NEXT:    [[TMP14:%.*]] = fcmp fast ogt float [[TMP12]], [[TMP13]]
; THRESH-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], float [[TMP12]], float [[TMP13]]
; THRESH-NEXT:    [[TMP16:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 5), align 4
; THRESH-NEXT:    [[TMP17:%.*]] = fcmp fast ogt float [[TMP15]], [[TMP16]]
; THRESH-NEXT:    [[TMP18:%.*]] = select i1 [[TMP17]], float [[TMP15]], float [[TMP16]]
; THRESH-NEXT:    [[TMP19:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 6), align 8
; THRESH-NEXT:    [[TMP20:%.*]] = fcmp fast ogt float [[TMP18]], [[TMP19]]
; THRESH-NEXT:    [[TMP21:%.*]] = select i1 [[TMP20]], float [[TMP18]], float [[TMP19]]
; THRESH-NEXT:    [[TMP22:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 7), align 4
; THRESH-NEXT:    [[TMP23:%.*]] = fcmp fast ogt float [[TMP21]], [[TMP22]]
; THRESH-NEXT:    [[TMP24:%.*]] = select i1 [[TMP23]], float [[TMP21]], float [[TMP22]]
; THRESH-NEXT:    [[TMP25:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 8), align 16
; THRESH-NEXT:    [[TMP26:%.*]] = fcmp fast ogt float [[TMP24]], [[TMP25]]
; THRESH-NEXT:    [[TMP27:%.*]] = select i1 [[TMP26]], float [[TMP24]], float [[TMP25]]
; THRESH-NEXT:    [[TMP28:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 9), align 4
; THRESH-NEXT:    [[TMP29:%.*]] = fcmp fast ogt float [[TMP27]], [[TMP28]]
; THRESH-NEXT:    [[TMP30:%.*]] = select i1 [[TMP29]], float [[TMP27]], float [[TMP28]]
; THRESH-NEXT:    [[TMP31:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 10), align 8
; THRESH-NEXT:    [[TMP32:%.*]] = fcmp fast ogt float [[TMP30]], [[TMP31]]
; THRESH-NEXT:    [[TMP33:%.*]] = select i1 [[TMP32]], float [[TMP30]], float [[TMP31]]
; THRESH-NEXT:    [[TMP34:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 11), align 4
; THRESH-NEXT:    [[TMP35:%.*]] = fcmp fast ogt float [[TMP33]], [[TMP34]]
; THRESH-NEXT:    [[TMP36:%.*]] = select i1 [[TMP35]], float [[TMP33]], float [[TMP34]]
; THRESH-NEXT:    [[TMP37:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 12), align 16
; THRESH-NEXT:    [[TMP38:%.*]] = fcmp fast ogt float [[TMP36]], [[TMP37]]
; THRESH-NEXT:    [[TMP39:%.*]] = select i1 [[TMP38]], float [[TMP36]], float [[TMP37]]
; THRESH-NEXT:    [[TMP40:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 13), align 4
; THRESH-NEXT:    [[TMP41:%.*]] = fcmp fast ogt float [[TMP39]], [[TMP40]]
; THRESH-NEXT:    [[TMP42:%.*]] = select i1 [[TMP41]], float [[TMP39]], float [[TMP40]]
; THRESH-NEXT:    [[TMP43:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 14), align 8
; THRESH-NEXT:    [[TMP44:%.*]] = fcmp fast ogt float [[TMP42]], [[TMP43]]
; THRESH-NEXT:    [[TMP45:%.*]] = select i1 [[TMP44]], float [[TMP42]], float [[TMP43]]
; THRESH-NEXT:    [[TMP46:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 15), align 4
; THRESH-NEXT:    [[TMP47:%.*]] = fcmp fast ogt float [[TMP45]], [[TMP46]]
; THRESH-NEXT:    [[TMP48:%.*]] = select i1 [[TMP47]], float [[TMP45]], float [[TMP46]]
; THRESH-NEXT:    ret float [[TMP48]]
;
  %2 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 0), align 16
  %3 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 1), align 4
  %4 = fcmp fast ogt float %2, %3
  %5 = select i1 %4, float %2, float %3
  %6 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 2), align 8
  %7 = fcmp fast ogt float %5, %6
  %8 = select i1 %7, float %5, float %6
  %9 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 3), align 4
  %10 = fcmp fast ogt float %8, %9
  %11 = select i1 %10, float %8, float %9
  %12 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 4), align 16
  %13 = fcmp fast ogt float %11, %12
  %14 = select i1 %13, float %11, float %12
  %15 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 5), align 4
  %16 = fcmp fast ogt float %14, %15
  %17 = select i1 %16, float %14, float %15
  %18 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 6), align 8
  %19 = fcmp fast ogt float %17, %18
  %20 = select i1 %19, float %17, float %18
  %21 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 7), align 4
  %22 = fcmp fast ogt float %20, %21
  %23 = select i1 %22, float %20, float %21
  %24 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 8), align 16
  %25 = fcmp fast ogt float %23, %24
  %26 = select i1 %25, float %23, float %24
  %27 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 9), align 4
  %28 = fcmp fast ogt float %26, %27
  %29 = select i1 %28, float %26, float %27
  %30 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 10), align 8
  %31 = fcmp fast ogt float %29, %30
  %32 = select i1 %31, float %29, float %30
  %33 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 11), align 4
  %34 = fcmp fast ogt float %32, %33
  %35 = select i1 %34, float %32, float %33
  %36 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 12), align 16
  %37 = fcmp fast ogt float %35, %36
  %38 = select i1 %37, float %35, float %36
  %39 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 13), align 4
  %40 = fcmp fast ogt float %38, %39
  %41 = select i1 %40, float %38, float %39
  %42 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 14), align 8
  %43 = fcmp fast ogt float %41, %42
  %44 = select i1 %43, float %41, float %42
  %45 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 15), align 4
  %46 = fcmp fast ogt float %44, %45
  %47 = select i1 %46, float %44, float %45
  ret float %47
}

; Note: legacy test - InstCombine creates maxnum intrinsics for fcmp+select with fastmath on the select.

define float @maxf32(float) {
; DEFAULT-LABEL: @maxf32(
; DEFAULT-NEXT:    [[TMP2:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 0), align 16
; DEFAULT-NEXT:    [[TMP3:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 1), align 4
; DEFAULT-NEXT:    [[TMP4:%.*]] = fcmp fast ogt float [[TMP2]], [[TMP3]]
; DEFAULT-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], float [[TMP2]], float [[TMP3]]
; DEFAULT-NEXT:    [[TMP6:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 2), align 8
; DEFAULT-NEXT:    [[TMP7:%.*]] = fcmp fast ogt float [[TMP5]], [[TMP6]]
; DEFAULT-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], float [[TMP5]], float [[TMP6]]
; DEFAULT-NEXT:    [[TMP9:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 3), align 4
; DEFAULT-NEXT:    [[TMP10:%.*]] = fcmp fast ogt float [[TMP8]], [[TMP9]]
; DEFAULT-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], float [[TMP8]], float [[TMP9]]
; DEFAULT-NEXT:    [[TMP12:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 4), align 16
; DEFAULT-NEXT:    [[TMP13:%.*]] = fcmp fast ogt float [[TMP11]], [[TMP12]]
; DEFAULT-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], float [[TMP11]], float [[TMP12]]
; DEFAULT-NEXT:    [[TMP15:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 5), align 4
; DEFAULT-NEXT:    [[TMP16:%.*]] = fcmp fast ogt float [[TMP14]], [[TMP15]]
; DEFAULT-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], float [[TMP14]], float [[TMP15]]
; DEFAULT-NEXT:    [[TMP18:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 6), align 8
; DEFAULT-NEXT:    [[TMP19:%.*]] = fcmp fast ogt float [[TMP17]], [[TMP18]]
; DEFAULT-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], float [[TMP17]], float [[TMP18]]
; DEFAULT-NEXT:    [[TMP21:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 7), align 4
; DEFAULT-NEXT:    [[TMP22:%.*]] = fcmp fast ogt float [[TMP20]], [[TMP21]]
; DEFAULT-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], float [[TMP20]], float [[TMP21]]
; DEFAULT-NEXT:    [[TMP24:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 8), align 16
; DEFAULT-NEXT:    [[TMP25:%.*]] = fcmp fast ogt float [[TMP23]], [[TMP24]]
; DEFAULT-NEXT:    [[TMP26:%.*]] = select i1 [[TMP25]], float [[TMP23]], float [[TMP24]]
; DEFAULT-NEXT:    [[TMP27:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 9), align 4
; DEFAULT-NEXT:    [[TMP28:%.*]] = fcmp fast ogt float [[TMP26]], [[TMP27]]
; DEFAULT-NEXT:    [[TMP29:%.*]] = select i1 [[TMP28]], float [[TMP26]], float [[TMP27]]
; DEFAULT-NEXT:    [[TMP30:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 10), align 8
; DEFAULT-NEXT:    [[TMP31:%.*]] = fcmp fast ogt float [[TMP29]], [[TMP30]]
; DEFAULT-NEXT:    [[TMP32:%.*]] = select i1 [[TMP31]], float [[TMP29]], float [[TMP30]]
; DEFAULT-NEXT:    [[TMP33:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 11), align 4
; DEFAULT-NEXT:    [[TMP34:%.*]] = fcmp fast ogt float [[TMP32]], [[TMP33]]
; DEFAULT-NEXT:    [[TMP35:%.*]] = select i1 [[TMP34]], float [[TMP32]], float [[TMP33]]
; DEFAULT-NEXT:    [[TMP36:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 12), align 16
; DEFAULT-NEXT:    [[TMP37:%.*]] = fcmp fast ogt float [[TMP35]], [[TMP36]]
; DEFAULT-NEXT:    [[TMP38:%.*]] = select i1 [[TMP37]], float [[TMP35]], float [[TMP36]]
; DEFAULT-NEXT:    [[TMP39:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 13), align 4
; DEFAULT-NEXT:    [[TMP40:%.*]] = fcmp fast ogt float [[TMP38]], [[TMP39]]
; DEFAULT-NEXT:    [[TMP41:%.*]] = select i1 [[TMP40]], float [[TMP38]], float [[TMP39]]
; DEFAULT-NEXT:    [[TMP42:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 14), align 8
; DEFAULT-NEXT:    [[TMP43:%.*]] = fcmp fast ogt float [[TMP41]], [[TMP42]]
; DEFAULT-NEXT:    [[TMP44:%.*]] = select i1 [[TMP43]], float [[TMP41]], float [[TMP42]]
; DEFAULT-NEXT:    [[TMP45:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 15), align 4
; DEFAULT-NEXT:    [[TMP46:%.*]] = fcmp fast ogt float [[TMP44]], [[TMP45]]
; DEFAULT-NEXT:    [[TMP47:%.*]] = select i1 [[TMP46]], float [[TMP44]], float [[TMP45]]
; DEFAULT-NEXT:    [[TMP48:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 16), align 16
; DEFAULT-NEXT:    [[TMP49:%.*]] = fcmp fast ogt float [[TMP47]], [[TMP48]]
; DEFAULT-NEXT:    [[TMP50:%.*]] = select i1 [[TMP49]], float [[TMP47]], float [[TMP48]]
; DEFAULT-NEXT:    [[TMP51:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 17), align 4
; DEFAULT-NEXT:    [[TMP52:%.*]] = fcmp fast ogt float [[TMP50]], [[TMP51]]
; DEFAULT-NEXT:    [[TMP53:%.*]] = select i1 [[TMP52]], float [[TMP50]], float [[TMP51]]
; DEFAULT-NEXT:    [[TMP54:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 18), align 8
; DEFAULT-NEXT:    [[TMP55:%.*]] = fcmp fast ogt float [[TMP53]], [[TMP54]]
; DEFAULT-NEXT:    [[TMP56:%.*]] = select i1 [[TMP55]], float [[TMP53]], float [[TMP54]]
; DEFAULT-NEXT:    [[TMP57:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 19), align 4
; DEFAULT-NEXT:    [[TMP58:%.*]] = fcmp fast ogt float [[TMP56]], [[TMP57]]
; DEFAULT-NEXT:    [[TMP59:%.*]] = select i1 [[TMP58]], float [[TMP56]], float [[TMP57]]
; DEFAULT-NEXT:    [[TMP60:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 20), align 16
; DEFAULT-NEXT:    [[TMP61:%.*]] = fcmp fast ogt float [[TMP59]], [[TMP60]]
; DEFAULT-NEXT:    [[TMP62:%.*]] = select i1 [[TMP61]], float [[TMP59]], float [[TMP60]]
; DEFAULT-NEXT:    [[TMP63:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 21), align 4
; DEFAULT-NEXT:    [[TMP64:%.*]] = fcmp fast ogt float [[TMP62]], [[TMP63]]
; DEFAULT-NEXT:    [[TMP65:%.*]] = select i1 [[TMP64]], float [[TMP62]], float [[TMP63]]
; DEFAULT-NEXT:    [[TMP66:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 22), align 8
; DEFAULT-NEXT:    [[TMP67:%.*]] = fcmp fast ogt float [[TMP65]], [[TMP66]]
; DEFAULT-NEXT:    [[TMP68:%.*]] = select i1 [[TMP67]], float [[TMP65]], float [[TMP66]]
; DEFAULT-NEXT:    [[TMP69:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 23), align 4
; DEFAULT-NEXT:    [[TMP70:%.*]] = fcmp fast ogt float [[TMP68]], [[TMP69]]
; DEFAULT-NEXT:    [[TMP71:%.*]] = select i1 [[TMP70]], float [[TMP68]], float [[TMP69]]
; DEFAULT-NEXT:    [[TMP72:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 24), align 16
; DEFAULT-NEXT:    [[TMP73:%.*]] = fcmp fast ogt float [[TMP71]], [[TMP72]]
; DEFAULT-NEXT:    [[TMP74:%.*]] = select i1 [[TMP73]], float [[TMP71]], float [[TMP72]]
; DEFAULT-NEXT:    [[TMP75:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 25), align 4
; DEFAULT-NEXT:    [[TMP76:%.*]] = fcmp fast ogt float [[TMP74]], [[TMP75]]
; DEFAULT-NEXT:    [[TMP77:%.*]] = select i1 [[TMP76]], float [[TMP74]], float [[TMP75]]
; DEFAULT-NEXT:    [[TMP78:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 26), align 8
; DEFAULT-NEXT:    [[TMP79:%.*]] = fcmp fast ogt float [[TMP77]], [[TMP78]]
; DEFAULT-NEXT:    [[TMP80:%.*]] = select i1 [[TMP79]], float [[TMP77]], float [[TMP78]]
; DEFAULT-NEXT:    [[TMP81:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 27), align 4
; DEFAULT-NEXT:    [[TMP82:%.*]] = fcmp fast ogt float [[TMP80]], [[TMP81]]
; DEFAULT-NEXT:    [[TMP83:%.*]] = select i1 [[TMP82]], float [[TMP80]], float [[TMP81]]
; DEFAULT-NEXT:    [[TMP84:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 28), align 16
; DEFAULT-NEXT:    [[TMP85:%.*]] = fcmp fast ogt float [[TMP83]], [[TMP84]]
; DEFAULT-NEXT:    [[TMP86:%.*]] = select i1 [[TMP85]], float [[TMP83]], float [[TMP84]]
; DEFAULT-NEXT:    [[TMP87:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 29), align 4
; DEFAULT-NEXT:    [[TMP88:%.*]] = fcmp fast ogt float [[TMP86]], [[TMP87]]
; DEFAULT-NEXT:    [[TMP89:%.*]] = select i1 [[TMP88]], float [[TMP86]], float [[TMP87]]
; DEFAULT-NEXT:    [[TMP90:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 30), align 8
; DEFAULT-NEXT:    [[TMP91:%.*]] = fcmp fast ogt float [[TMP89]], [[TMP90]]
; DEFAULT-NEXT:    [[TMP92:%.*]] = select i1 [[TMP91]], float [[TMP89]], float [[TMP90]]
; DEFAULT-NEXT:    [[TMP93:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 31), align 4
; DEFAULT-NEXT:    [[TMP94:%.*]] = fcmp fast ogt float [[TMP92]], [[TMP93]]
; DEFAULT-NEXT:    [[TMP95:%.*]] = select i1 [[TMP94]], float [[TMP92]], float [[TMP93]]
; DEFAULT-NEXT:    ret float [[TMP95]]
;
; THRESH-LABEL: @maxf32(
; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x float>, <2 x float>* bitcast ([32 x float]* @arr1 to <2 x float>*), align 16
; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[TMP2]], i32 0
; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x float> [[TMP2]], i32 1
; THRESH-NEXT:    [[TMP5:%.*]] = fcmp fast ogt float [[TMP3]], [[TMP4]]
; THRESH-NEXT:    [[TMP6:%.*]] = select i1 [[TMP5]], float [[TMP3]], float [[TMP4]]
; THRESH-NEXT:    [[TMP7:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 2), align 8
; THRESH-NEXT:    [[TMP8:%.*]] = fcmp fast ogt float [[TMP6]], [[TMP7]]
; THRESH-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], float [[TMP6]], float [[TMP7]]
; THRESH-NEXT:    [[TMP10:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 3), align 4
; THRESH-NEXT:    [[TMP11:%.*]] = fcmp fast ogt float [[TMP9]], [[TMP10]]
; THRESH-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], float [[TMP9]], float [[TMP10]]
; THRESH-NEXT:    [[TMP13:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 4), align 16
; THRESH-NEXT:    [[TMP14:%.*]] = fcmp fast ogt float [[TMP12]], [[TMP13]]
; THRESH-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], float [[TMP12]], float [[TMP13]]
; THRESH-NEXT:    [[TMP16:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 5), align 4
; THRESH-NEXT:    [[TMP17:%.*]] = fcmp fast ogt float [[TMP15]], [[TMP16]]
; THRESH-NEXT:    [[TMP18:%.*]] = select i1 [[TMP17]], float [[TMP15]], float [[TMP16]]
; THRESH-NEXT:    [[TMP19:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 6), align 8
; THRESH-NEXT:    [[TMP20:%.*]] = fcmp fast ogt float [[TMP18]], [[TMP19]]
; THRESH-NEXT:    [[TMP21:%.*]] = select i1 [[TMP20]], float [[TMP18]], float [[TMP19]]
; THRESH-NEXT:    [[TMP22:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 7), align 4
; THRESH-NEXT:    [[TMP23:%.*]] = fcmp fast ogt float [[TMP21]], [[TMP22]]
; THRESH-NEXT:    [[TMP24:%.*]] = select i1 [[TMP23]], float [[TMP21]], float [[TMP22]]
; THRESH-NEXT:    [[TMP25:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 8), align 16
; THRESH-NEXT:    [[TMP26:%.*]] = fcmp fast ogt float [[TMP24]], [[TMP25]]
; THRESH-NEXT:    [[TMP27:%.*]] = select i1 [[TMP26]], float [[TMP24]], float [[TMP25]]
; THRESH-NEXT:    [[TMP28:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 9), align 4
; THRESH-NEXT:    [[TMP29:%.*]] = fcmp fast ogt float [[TMP27]], [[TMP28]]
; THRESH-NEXT:    [[TMP30:%.*]] = select i1 [[TMP29]], float [[TMP27]], float [[TMP28]]
; THRESH-NEXT:    [[TMP31:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 10), align 8
; THRESH-NEXT:    [[TMP32:%.*]] = fcmp fast ogt float [[TMP30]], [[TMP31]]
; THRESH-NEXT:    [[TMP33:%.*]] = select i1 [[TMP32]], float [[TMP30]], float [[TMP31]]
; THRESH-NEXT:    [[TMP34:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 11), align 4
; THRESH-NEXT:    [[TMP35:%.*]] = fcmp fast ogt float [[TMP33]], [[TMP34]]
; THRESH-NEXT:    [[TMP36:%.*]] = select i1 [[TMP35]], float [[TMP33]], float [[TMP34]]
; THRESH-NEXT:    [[TMP37:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 12), align 16
; THRESH-NEXT:    [[TMP38:%.*]] = fcmp fast ogt float [[TMP36]], [[TMP37]]
; THRESH-NEXT:    [[TMP39:%.*]] = select i1 [[TMP38]], float [[TMP36]], float [[TMP37]]
; THRESH-NEXT:    [[TMP40:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 13), align 4
; THRESH-NEXT:    [[TMP41:%.*]] = fcmp fast ogt float [[TMP39]], [[TMP40]]
; THRESH-NEXT:    [[TMP42:%.*]] = select i1 [[TMP41]], float [[TMP39]], float [[TMP40]]
; THRESH-NEXT:    [[TMP43:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 14), align 8
; THRESH-NEXT:    [[TMP44:%.*]] = fcmp fast ogt float [[TMP42]], [[TMP43]]
; THRESH-NEXT:    [[TMP45:%.*]] = select i1 [[TMP44]], float [[TMP42]], float [[TMP43]]
; THRESH-NEXT:    [[TMP46:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 15), align 4
; THRESH-NEXT:    [[TMP47:%.*]] = fcmp fast ogt float [[TMP45]], [[TMP46]]
; THRESH-NEXT:    [[TMP48:%.*]] = select i1 [[TMP47]], float [[TMP45]], float [[TMP46]]
; THRESH-NEXT:    [[TMP49:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 16), align 16
; THRESH-NEXT:    [[TMP50:%.*]] = fcmp fast ogt float [[TMP48]], [[TMP49]]
; THRESH-NEXT:    [[TMP51:%.*]] = select i1 [[TMP50]], float [[TMP48]], float [[TMP49]]
; THRESH-NEXT:    [[TMP52:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 17), align 4
; THRESH-NEXT:    [[TMP53:%.*]] = fcmp fast ogt float [[TMP51]], [[TMP52]]
; THRESH-NEXT:    [[TMP54:%.*]] = select i1 [[TMP53]], float [[TMP51]], float [[TMP52]]
; THRESH-NEXT:    [[TMP55:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 18), align 8
; THRESH-NEXT:    [[TMP56:%.*]] = fcmp fast ogt float [[TMP54]], [[TMP55]]
; THRESH-NEXT:    [[TMP57:%.*]] = select i1 [[TMP56]], float [[TMP54]], float [[TMP55]]
; THRESH-NEXT:    [[TMP58:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 19), align 4
; THRESH-NEXT:    [[TMP59:%.*]] = fcmp fast ogt float [[TMP57]], [[TMP58]]
; THRESH-NEXT:    [[TMP60:%.*]] = select i1 [[TMP59]], float [[TMP57]], float [[TMP58]]
; THRESH-NEXT:    [[TMP61:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 20), align 16
; THRESH-NEXT:    [[TMP62:%.*]] = fcmp fast ogt float [[TMP60]], [[TMP61]]
; THRESH-NEXT:    [[TMP63:%.*]] = select i1 [[TMP62]], float [[TMP60]], float [[TMP61]]
; THRESH-NEXT:    [[TMP64:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 21), align 4
; THRESH-NEXT:    [[TMP65:%.*]] = fcmp fast ogt float [[TMP63]], [[TMP64]]
; THRESH-NEXT:    [[TMP66:%.*]] = select i1 [[TMP65]], float [[TMP63]], float [[TMP64]]
; THRESH-NEXT:    [[TMP67:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 22), align 8
; THRESH-NEXT:    [[TMP68:%.*]] = fcmp fast ogt float [[TMP66]], [[TMP67]]
; THRESH-NEXT:    [[TMP69:%.*]] = select i1 [[TMP68]], float [[TMP66]], float [[TMP67]]
; THRESH-NEXT:    [[TMP70:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 23), align 4
; THRESH-NEXT:    [[TMP71:%.*]] = fcmp fast ogt float [[TMP69]], [[TMP70]]
; THRESH-NEXT:    [[TMP72:%.*]] = select i1 [[TMP71]], float [[TMP69]], float [[TMP70]]
; THRESH-NEXT:    [[TMP73:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 24), align 16
; THRESH-NEXT:    [[TMP74:%.*]] = fcmp fast ogt float [[TMP72]], [[TMP73]]
; THRESH-NEXT:    [[TMP75:%.*]] = select i1 [[TMP74]], float [[TMP72]], float [[TMP73]]
; THRESH-NEXT:    [[TMP76:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 25), align 4
; THRESH-NEXT:    [[TMP77:%.*]] = fcmp fast ogt float [[TMP75]], [[TMP76]]
; THRESH-NEXT:    [[TMP78:%.*]] = select i1 [[TMP77]], float [[TMP75]], float [[TMP76]]
; THRESH-NEXT:    [[TMP79:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 26), align 8
; THRESH-NEXT:    [[TMP80:%.*]] = fcmp fast ogt float [[TMP78]], [[TMP79]]
; THRESH-NEXT:    [[TMP81:%.*]] = select i1 [[TMP80]], float [[TMP78]], float [[TMP79]]
; THRESH-NEXT:    [[TMP82:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 27), align 4
; THRESH-NEXT:    [[TMP83:%.*]] = fcmp fast ogt float [[TMP81]], [[TMP82]]
; THRESH-NEXT:    [[TMP84:%.*]] = select i1 [[TMP83]], float [[TMP81]], float [[TMP82]]
; THRESH-NEXT:    [[TMP85:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 28), align 16
; THRESH-NEXT:    [[TMP86:%.*]] = fcmp fast ogt float [[TMP84]], [[TMP85]]
; THRESH-NEXT:    [[TMP87:%.*]] = select i1 [[TMP86]], float [[TMP84]], float [[TMP85]]
; THRESH-NEXT:    [[TMP88:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 29), align 4
; THRESH-NEXT:    [[TMP89:%.*]] = fcmp fast ogt float [[TMP87]], [[TMP88]]
; THRESH-NEXT:    [[TMP90:%.*]] = select i1 [[TMP89]], float [[TMP87]], float [[TMP88]]
; THRESH-NEXT:    [[TMP91:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 30), align 8
; THRESH-NEXT:    [[TMP92:%.*]] = fcmp fast ogt float [[TMP90]], [[TMP91]]
; THRESH-NEXT:    [[TMP93:%.*]] = select i1 [[TMP92]], float [[TMP90]], float [[TMP91]]
; THRESH-NEXT:    [[TMP94:%.*]] = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 31), align 4
; THRESH-NEXT:    [[TMP95:%.*]] = fcmp fast ogt float [[TMP93]], [[TMP94]]
; THRESH-NEXT:    [[TMP96:%.*]] = select i1 [[TMP95]], float [[TMP93]], float [[TMP94]]
; THRESH-NEXT:    ret float [[TMP96]]
;
  %2 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 0), align 16
  %3 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 1), align 4
  %4 = fcmp fast ogt float %2, %3
  %5 = select i1 %4, float %2, float %3
  %6 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 2), align 8
  %7 = fcmp fast ogt float %5, %6
  %8 = select i1 %7, float %5, float %6
  %9 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 3), align 4
  %10 = fcmp fast ogt float %8, %9
  %11 = select i1 %10, float %8, float %9
  %12 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 4), align 16
  %13 = fcmp fast ogt float %11, %12
  %14 = select i1 %13, float %11, float %12
  %15 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 5), align 4
  %16 = fcmp fast ogt float %14, %15
  %17 = select i1 %16, float %14, float %15
  %18 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 6), align 8
  %19 = fcmp fast ogt float %17, %18
  %20 = select i1 %19, float %17, float %18
  %21 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 7), align 4
  %22 = fcmp fast ogt float %20, %21
  %23 = select i1 %22, float %20, float %21
  %24 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 8), align 16
  %25 = fcmp fast ogt float %23, %24
  %26 = select i1 %25, float %23, float %24
  %27 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 9), align 4
  %28 = fcmp fast ogt float %26, %27
  %29 = select i1 %28, float %26, float %27
  %30 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 10), align 8
  %31 = fcmp fast ogt float %29, %30
  %32 = select i1 %31, float %29, float %30
  %33 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 11), align 4
  %34 = fcmp fast ogt float %32, %33
  %35 = select i1 %34, float %32, float %33
  %36 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 12), align 16
  %37 = fcmp fast ogt float %35, %36
  %38 = select i1 %37, float %35, float %36
  %39 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 13), align 4
  %40 = fcmp fast ogt float %38, %39
  %41 = select i1 %40, float %38, float %39
  %42 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 14), align 8
  %43 = fcmp fast ogt float %41, %42
  %44 = select i1 %43, float %41, float %42
  %45 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 15), align 4
  %46 = fcmp fast ogt float %44, %45
  %47 = select i1 %46, float %44, float %45
  %48 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 16), align 16
  %49 = fcmp fast ogt float %47, %48
  %50 = select i1 %49, float %47, float %48
  %51 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 17), align 4
  %52 = fcmp fast ogt float %50, %51
  %53 = select i1 %52, float %50, float %51
  %54 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 18), align 8
  %55 = fcmp fast ogt float %53, %54
  %56 = select i1 %55, float %53, float %54
  %57 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 19), align 4
  %58 = fcmp fast ogt float %56, %57
  %59 = select i1 %58, float %56, float %57
  %60 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 20), align 16
  %61 = fcmp fast ogt float %59, %60
  %62 = select i1 %61, float %59, float %60
  %63 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 21), align 4
  %64 = fcmp fast ogt float %62, %63
  %65 = select i1 %64, float %62, float %63
  %66 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 22), align 8
  %67 = fcmp fast ogt float %65, %66
  %68 = select i1 %67, float %65, float %66
  %69 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 23), align 4
  %70 = fcmp fast ogt float %68, %69
  %71 = select i1 %70, float %68, float %69
  %72 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 24), align 16
  %73 = fcmp fast ogt float %71, %72
  %74 = select i1 %73, float %71, float %72
  %75 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 25), align 4
  %76 = fcmp fast ogt float %74, %75
  %77 = select i1 %76, float %74, float %75
  %78 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 26), align 8
  %79 = fcmp fast ogt float %77, %78
  %80 = select i1 %79, float %77, float %78
  %81 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 27), align 4
  %82 = fcmp fast ogt float %80, %81
  %83 = select i1 %82, float %80, float %81
  %84 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 28), align 16
  %85 = fcmp fast ogt float %83, %84
  %86 = select i1 %85, float %83, float %84
  %87 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 29), align 4
  %88 = fcmp fast ogt float %86, %87
  %89 = select i1 %88, float %86, float %87
  %90 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 30), align 8
  %91 = fcmp fast ogt float %89, %90
  %92 = select i1 %91, float %89, float %90
  %93 = load float, float* getelementptr inbounds ([32 x float], [32 x float]* @arr1, i64 0, i64 31), align 4
  %94 = fcmp fast ogt float %92, %93
  %95 = select i1 %94, float %92, float %93
  ret float %95
}

define i32 @maxi8_mutiple_uses(i32) {
; SSE-LABEL: @maxi8_mutiple_uses(
; SSE-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
; SSE-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
; SSE-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
; SSE-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
; SSE-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
; SSE-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
; SSE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 [[TMP5]], i32 [[TMP6]]
; SSE-NEXT:    [[TMP9:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
; SSE-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP8]], i32 [[TMP9]]
; SSE-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
; SSE-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
; SSE-NEXT:    [[TMP15:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
; SSE-NEXT:    [[TMP16:%.*]] = icmp sgt i32 [[TMP14]], [[TMP15]]
; SSE-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP14]], i32 [[TMP15]]
; SSE-NEXT:    [[TMP18:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; SSE-NEXT:    [[TMP19:%.*]] = icmp sgt i32 [[TMP17]], [[TMP18]]
; SSE-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP17]], i32 [[TMP18]]
; SSE-NEXT:    [[TMP21:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; SSE-NEXT:    [[TMP22:%.*]] = icmp sgt i32 [[TMP20]], [[TMP21]]
; SSE-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP20]], i32 [[TMP21]]
; SSE-NEXT:    [[TMP24:%.*]] = select i1 [[TMP4]], i32 3, i32 4
; SSE-NEXT:    store i32 [[TMP24]], i32* @var, align 8
; SSE-NEXT:    ret i32 [[TMP23]]
;
; AVX-LABEL: @maxi8_mutiple_uses(
; AVX-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
; AVX-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
; AVX-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
; AVX-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
; AVX-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
; AVX-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; AVX-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; AVX-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP6]])
; AVX-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
; AVX-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
; AVX-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
; AVX-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 [[TMP8]]
; AVX-NEXT:    [[OP_EXTRA:%.*]] = icmp sgt i32 [[TMP13]], [[TMP5]]
; AVX-NEXT:    [[OP_EXTRA1:%.*]] = select i1 [[OP_EXTRA]], i32 [[TMP13]], i32 [[TMP5]]
; AVX-NEXT:    [[TMP14:%.*]] = select i1 [[TMP4]], i32 3, i32 4
; AVX-NEXT:    store i32 [[TMP14]], i32* @var, align 8
; AVX-NEXT:    ret i32 [[OP_EXTRA1]]
;
; AVX2-LABEL: @maxi8_mutiple_uses(
; AVX2-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
; AVX2-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
; AVX2-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
; AVX2-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
; AVX2-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
; AVX2-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; AVX2-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; AVX2-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP6]])
; AVX2-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
; AVX2-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
; AVX2-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
; AVX2-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 [[TMP8]]
; AVX2-NEXT:    [[OP_EXTRA:%.*]] = icmp sgt i32 [[TMP13]], [[TMP5]]
; AVX2-NEXT:    [[OP_EXTRA1:%.*]] = select i1 [[OP_EXTRA]], i32 [[TMP13]], i32 [[TMP5]]
; AVX2-NEXT:    [[TMP14:%.*]] = select i1 [[TMP4]], i32 3, i32 4
; AVX2-NEXT:    store i32 [[TMP14]], i32* @var, align 8
; AVX2-NEXT:    ret i32 [[OP_EXTRA1]]
;
; THRESH-LABEL: @maxi8_mutiple_uses(
; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* bitcast ([32 x i32]* @arr to <2 x i32>*), align 16
; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 0
; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
; THRESH-NEXT:    [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
; THRESH-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; THRESH-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; THRESH-NEXT:    [[TMP8:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP5]])
; THRESH-NEXT:    [[TMP9:%.*]] = icmp sgt i32 [[TMP8]], [[TMP6]]
; THRESH-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP8]], i32 [[TMP6]]
; THRESH-NEXT:    [[TMP11:%.*]] = insertelement <2 x i32> poison, i32 [[TMP10]], i32 0
; THRESH-NEXT:    [[TMP12:%.*]] = insertelement <2 x i32> [[TMP11]], i32 [[TMP3]], i32 1
; THRESH-NEXT:    [[TMP13:%.*]] = insertelement <2 x i32> poison, i32 [[TMP7]], i32 0
; THRESH-NEXT:    [[TMP14:%.*]] = insertelement <2 x i32> [[TMP13]], i32 [[TMP4]], i32 1
; THRESH-NEXT:    [[TMP15:%.*]] = icmp sgt <2 x i32> [[TMP12]], [[TMP14]]
; THRESH-NEXT:    [[TMP16:%.*]] = select <2 x i1> [[TMP15]], <2 x i32> [[TMP12]], <2 x i32> [[TMP14]]
; THRESH-NEXT:    [[TMP17:%.*]] = extractelement <2 x i32> [[TMP16]], i32 0
; THRESH-NEXT:    [[TMP18:%.*]] = extractelement <2 x i32> [[TMP16]], i32 1
; THRESH-NEXT:    [[OP_EXTRA:%.*]] = icmp sgt i32 [[TMP17]], [[TMP18]]
; THRESH-NEXT:    [[OP_EXTRA1:%.*]] = select i1 [[OP_EXTRA]], i32 [[TMP17]], i32 [[TMP18]]
; THRESH-NEXT:    [[TMP19:%.*]] = extractelement <2 x i1> [[TMP15]], i32 1
; THRESH-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32 3, i32 4
; THRESH-NEXT:    store i32 [[TMP20]], i32* @var, align 8
; THRESH-NEXT:    ret i32 [[OP_EXTRA1]]
;
  %2 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
  %3 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
  %4 = icmp sgt i32 %2, %3
  %5 = select i1 %4, i32 %2, i32 %3
  %6 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
  %7 = icmp sgt i32 %5, %6
  %8 = select i1 %7, i32 %5, i32 %6
  %9 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
  %10 = icmp sgt i32 %8, %9
  %11 = select i1 %10, i32 %8, i32 %9
  %12 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
  %13 = icmp sgt i32 %11, %12
  %14 = select i1 %13, i32 %11, i32 %12
  %15 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
  %16 = icmp sgt i32 %14, %15
  %17 = select i1 %16, i32 %14, i32 %15
  %18 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
  %19 = icmp sgt i32 %17, %18
  %20 = select i1 %19, i32 %17, i32 %18
  %21 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
  %22 = icmp sgt i32 %20, %21
  %23 = select i1 %22, i32 %20, i32 %21
  %24 = select i1 %4, i32 3, i32 4
  store i32 %24, i32* @var, align 8
  ret i32 %23
}

define i32 @maxi8_mutiple_uses2(i32) {
; DEFAULT-LABEL: @maxi8_mutiple_uses2(
; DEFAULT-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
; DEFAULT-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
; DEFAULT-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
; DEFAULT-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
; DEFAULT-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
; DEFAULT-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
; DEFAULT-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 [[TMP5]], i32 [[TMP6]]
; DEFAULT-NEXT:    [[TMP9:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
; DEFAULT-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
; DEFAULT-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP8]], i32 [[TMP9]]
; DEFAULT-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
; DEFAULT-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
; DEFAULT-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
; DEFAULT-NEXT:    [[TMP15:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
; DEFAULT-NEXT:    [[TMP16:%.*]] = icmp sgt i32 [[TMP14]], [[TMP15]]
; DEFAULT-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP14]], i32 [[TMP15]]
; DEFAULT-NEXT:    [[TMP18:%.*]] = select i1 [[TMP10]], i32 3, i32 4
; DEFAULT-NEXT:    store i32 [[TMP18]], i32* @var, align 8
; DEFAULT-NEXT:    ret i32 [[TMP17]]
;
; THRESH-LABEL: @maxi8_mutiple_uses2(
; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* bitcast ([32 x i32]* @arr to <2 x i32>*), align 16
; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 0
; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
; THRESH-NEXT:    [[TMP5:%.*]] = icmp sgt i32 [[TMP3]], [[TMP4]]
; THRESH-NEXT:    [[TMP6:%.*]] = select i1 [[TMP5]], i32 [[TMP3]], i32 [[TMP4]]
; THRESH-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
; THRESH-NEXT:    [[TMP8:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
; THRESH-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i32 [[TMP6]], i32 [[TMP7]]
; THRESH-NEXT:    [[TMP10:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
; THRESH-NEXT:    [[TMP11:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
; THRESH-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i32 [[TMP9]], i32 [[TMP10]]
; THRESH-NEXT:    [[TMP13:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
; THRESH-NEXT:    [[TMP14:%.*]] = icmp sgt i32 [[TMP12]], [[TMP13]]
; THRESH-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP12]], i32 [[TMP13]]
; THRESH-NEXT:    [[TMP16:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
; THRESH-NEXT:    [[TMP17:%.*]] = icmp sgt i32 [[TMP15]], [[TMP16]]
; THRESH-NEXT:    [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP15]], i32 [[TMP16]]
; THRESH-NEXT:    [[TMP19:%.*]] = select i1 [[TMP11]], i32 3, i32 4
; THRESH-NEXT:    store i32 [[TMP19]], i32* @var, align 8
; THRESH-NEXT:    ret i32 [[TMP18]]
;
  %2 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
  %3 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
  %4 = icmp sgt i32 %2, %3
  %5 = select i1 %4, i32 %2, i32 %3
  %6 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
  %7 = icmp sgt i32 %5, %6
  %8 = select i1 %7, i32 %5, i32 %6
  %9 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
  %10 = icmp sgt i32 %8, %9
  %11 = select i1 %10, i32 %8, i32 %9
  %12 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
  %13 = icmp sgt i32 %11, %12
  %14 = select i1 %13, i32 %11, i32 %12
  %15 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
  %16 = icmp sgt i32 %14, %15
  %17 = select i1 %16, i32 %14, i32 %15
  %18 = select i1 %10, i32 3, i32 4
  store i32 %18, i32* @var, align 8
  ret i32 %17
}

define i32 @maxi8_wrong_parent(i32) {
; SSE-LABEL: @maxi8_wrong_parent(
; SSE-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
; SSE-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
; SSE-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
; SSE-NEXT:    br label [[PP:%.*]]
; SSE:       pp:
; SSE-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
; SSE-NEXT:    [[TMP6:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
; SSE-NEXT:    [[TMP7:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
; SSE-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32 [[TMP5]], i32 [[TMP6]]
; SSE-NEXT:    [[TMP9:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
; SSE-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
; SSE-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP8]], i32 [[TMP9]]
; SSE-NEXT:    [[TMP12:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
; SSE-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
; SSE-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
; SSE-NEXT:    [[TMP15:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
; SSE-NEXT:    [[TMP16:%.*]] = icmp sgt i32 [[TMP14]], [[TMP15]]
; SSE-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP14]], i32 [[TMP15]]
; SSE-NEXT:    [[TMP18:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; SSE-NEXT:    [[TMP19:%.*]] = icmp sgt i32 [[TMP17]], [[TMP18]]
; SSE-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP17]], i32 [[TMP18]]
; SSE-NEXT:    [[TMP21:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; SSE-NEXT:    [[TMP22:%.*]] = icmp sgt i32 [[TMP20]], [[TMP21]]
; SSE-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP20]], i32 [[TMP21]]
; SSE-NEXT:    ret i32 [[TMP23]]
;
; AVX-LABEL: @maxi8_wrong_parent(
; AVX-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
; AVX-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
; AVX-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
; AVX-NEXT:    br label [[PP:%.*]]
; AVX:       pp:
; AVX-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
; AVX-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
; AVX-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; AVX-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; AVX-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP6]])
; AVX-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
; AVX-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
; AVX-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
; AVX-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 [[TMP8]]
; AVX-NEXT:    [[OP_EXTRA:%.*]] = icmp sgt i32 [[TMP13]], [[TMP5]]
; AVX-NEXT:    [[OP_EXTRA1:%.*]] = select i1 [[OP_EXTRA]], i32 [[TMP13]], i32 [[TMP5]]
; AVX-NEXT:    ret i32 [[OP_EXTRA1]]
;
; AVX2-LABEL: @maxi8_wrong_parent(
; AVX2-NEXT:    [[TMP2:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
; AVX2-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
; AVX2-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP2]], [[TMP3]]
; AVX2-NEXT:    br label [[PP:%.*]]
; AVX2:       pp:
; AVX2-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32 [[TMP2]], i32 [[TMP3]]
; AVX2-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
; AVX2-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; AVX2-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; AVX2-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP6]])
; AVX2-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
; AVX2-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
; AVX2-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
; AVX2-NEXT:    [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP11]], i32 [[TMP8]]
; AVX2-NEXT:    [[OP_EXTRA:%.*]] = icmp sgt i32 [[TMP13]], [[TMP5]]
; AVX2-NEXT:    [[OP_EXTRA1:%.*]] = select i1 [[OP_EXTRA]], i32 [[TMP13]], i32 [[TMP5]]
; AVX2-NEXT:    ret i32 [[OP_EXTRA1]]
;
; THRESH-LABEL: @maxi8_wrong_parent(
; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* bitcast ([32 x i32]* @arr to <2 x i32>*), align 16
; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[TMP2]], i32 0
; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32> [[TMP2]], i32 1
; THRESH-NEXT:    [[TMP5:%.*]] = icmp sgt i32 [[TMP3]], [[TMP4]]
; THRESH-NEXT:    br label [[PP:%.*]]
; THRESH:       pp:
; THRESH-NEXT:    [[TMP6:%.*]] = load <4 x i32>, <4 x i32>* bitcast (i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2) to <4 x i32>*), align 8
; THRESH-NEXT:    [[TMP7:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
; THRESH-NEXT:    [[TMP8:%.*]] = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
; THRESH-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP6]])
; THRESH-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP9]], [[TMP7]]
; THRESH-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32 [[TMP9]], i32 [[TMP7]]
; THRESH-NEXT:    [[TMP12:%.*]] = icmp sgt i32 [[TMP11]], [[TMP8]]
; THRESH-NEXT:    [[TMP13:%.*]] = insertelement <2 x i1> poison, i1 [[TMP12]], i32 0
; THRESH-NEXT:    [[TMP14:%.*]] = insertelement <2 x i1> [[TMP13]], i1 [[TMP5]], i32 1
; THRESH-NEXT:    [[TMP15:%.*]] = insertelement <2 x i32> poison, i32 [[TMP11]], i32 0
; THRESH-NEXT:    [[TMP16:%.*]] = insertelement <2 x i32> [[TMP15]], i32 [[TMP3]], i32 1
; THRESH-NEXT:    [[TMP17:%.*]] = insertelement <2 x i32> poison, i32 [[TMP8]], i32 0
; THRESH-NEXT:    [[TMP18:%.*]] = insertelement <2 x i32> [[TMP17]], i32 [[TMP4]], i32 1
; THRESH-NEXT:    [[TMP19:%.*]] = select <2 x i1> [[TMP14]], <2 x i32> [[TMP16]], <2 x i32> [[TMP18]]
; THRESH-NEXT:    [[TMP20:%.*]] = extractelement <2 x i32> [[TMP19]], i32 0
; THRESH-NEXT:    [[TMP21:%.*]] = extractelement <2 x i32> [[TMP19]], i32 1
; THRESH-NEXT:    [[OP_EXTRA:%.*]] = icmp sgt i32 [[TMP20]], [[TMP21]]
; THRESH-NEXT:    [[OP_EXTRA1:%.*]] = select i1 [[OP_EXTRA]], i32 [[TMP20]], i32 [[TMP21]]
; THRESH-NEXT:    ret i32 [[OP_EXTRA1]]
;
  %2 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 0), align 16
  %3 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 1), align 4
  %4 = icmp sgt i32 %2, %3
  br label %pp

pp:
  %5 = select i1 %4, i32 %2, i32 %3
  %6 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 2), align 8
  %7 = icmp sgt i32 %5, %6
  %8 = select i1 %7, i32 %5, i32 %6
  %9 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 3), align 4
  %10 = icmp sgt i32 %8, %9
  %11 = select i1 %10, i32 %8, i32 %9
  %12 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 4), align 16
  %13 = icmp sgt i32 %11, %12
  %14 = select i1 %13, i32 %11, i32 %12
  %15 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 5), align 4
  %16 = icmp sgt i32 %14, %15
  %17 = select i1 %16, i32 %14, i32 %15
  %18 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 6), align 8
  %19 = icmp sgt i32 %17, %18
  %20 = select i1 %19, i32 %17, i32 %18
  %21 = load i32, i32* getelementptr inbounds ([32 x i32], [32 x i32]* @arr, i64 0, i64 7), align 4
  %22 = icmp sgt i32 %20, %21
  %23 = select i1 %22, i32 %20, i32 %21
  ret i32 %23
}

; PR38191 - We don't handle array-of-pointer reductions.
define i32* @maxp8(i32) {
; DEFAULT-LABEL: @maxp8(
; DEFAULT-NEXT:    [[TMP2:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 0), align 16
; DEFAULT-NEXT:    [[TMP3:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 1), align 4
; DEFAULT-NEXT:    [[TMP4:%.*]] = icmp ugt i32* [[TMP2]], [[TMP3]]
; DEFAULT-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i32* [[TMP2]], i32* [[TMP3]]
; DEFAULT-NEXT:    [[TMP6:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 2), align 8
; DEFAULT-NEXT:    [[TMP7:%.*]] = icmp ugt i32* [[TMP5]], [[TMP6]]
; DEFAULT-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i32* [[TMP5]], i32* [[TMP6]]
; DEFAULT-NEXT:    [[TMP9:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 3), align 4
; DEFAULT-NEXT:    [[TMP10:%.*]] = icmp ugt i32* [[TMP8]], [[TMP9]]
; DEFAULT-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i32* [[TMP8]], i32* [[TMP9]]
; DEFAULT-NEXT:    [[TMP12:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 4), align 16
; DEFAULT-NEXT:    [[TMP13:%.*]] = icmp ugt i32* [[TMP11]], [[TMP12]]
; DEFAULT-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32* [[TMP11]], i32* [[TMP12]]
; DEFAULT-NEXT:    [[TMP15:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 5), align 4
; DEFAULT-NEXT:    [[TMP16:%.*]] = icmp ugt i32* [[TMP14]], [[TMP15]]
; DEFAULT-NEXT:    [[TMP17:%.*]] = select i1 [[TMP16]], i32* [[TMP14]], i32* [[TMP15]]
; DEFAULT-NEXT:    [[TMP18:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 6), align 8
; DEFAULT-NEXT:    [[TMP19:%.*]] = icmp ugt i32* [[TMP17]], [[TMP18]]
; DEFAULT-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i32* [[TMP17]], i32* [[TMP18]]
; DEFAULT-NEXT:    [[TMP21:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 7), align 4
; DEFAULT-NEXT:    [[TMP22:%.*]] = icmp ugt i32* [[TMP20]], [[TMP21]]
; DEFAULT-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i32* [[TMP20]], i32* [[TMP21]]
; DEFAULT-NEXT:    ret i32* [[TMP23]]
;
; THRESH-LABEL: @maxp8(
; THRESH-NEXT:    [[TMP2:%.*]] = load <2 x i32*>, <2 x i32*>* bitcast ([32 x i32*]* @arrp to <2 x i32*>*), align 16
; THRESH-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32*> [[TMP2]], i32 0
; THRESH-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32*> [[TMP2]], i32 1
; THRESH-NEXT:    [[TMP5:%.*]] = icmp ugt i32* [[TMP3]], [[TMP4]]
; THRESH-NEXT:    [[TMP6:%.*]] = select i1 [[TMP5]], i32* [[TMP3]], i32* [[TMP4]]
; THRESH-NEXT:    [[TMP7:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 2), align 8
; THRESH-NEXT:    [[TMP8:%.*]] = icmp ugt i32* [[TMP6]], [[TMP7]]
; THRESH-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i32* [[TMP6]], i32* [[TMP7]]
; THRESH-NEXT:    [[TMP10:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 3), align 4
; THRESH-NEXT:    [[TMP11:%.*]] = icmp ugt i32* [[TMP9]], [[TMP10]]
; THRESH-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i32* [[TMP9]], i32* [[TMP10]]
; THRESH-NEXT:    [[TMP13:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 4), align 16
; THRESH-NEXT:    [[TMP14:%.*]] = icmp ugt i32* [[TMP12]], [[TMP13]]
; THRESH-NEXT:    [[TMP15:%.*]] = select i1 [[TMP14]], i32* [[TMP12]], i32* [[TMP13]]
; THRESH-NEXT:    [[TMP16:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 5), align 4
; THRESH-NEXT:    [[TMP17:%.*]] = icmp ugt i32* [[TMP15]], [[TMP16]]
; THRESH-NEXT:    [[TMP18:%.*]] = select i1 [[TMP17]], i32* [[TMP15]], i32* [[TMP16]]
; THRESH-NEXT:    [[TMP19:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 6), align 8
; THRESH-NEXT:    [[TMP20:%.*]] = icmp ugt i32* [[TMP18]], [[TMP19]]
; THRESH-NEXT:    [[TMP21:%.*]] = select i1 [[TMP20]], i32* [[TMP18]], i32* [[TMP19]]
; THRESH-NEXT:    [[TMP22:%.*]] = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 7), align 4
; THRESH-NEXT:    [[TMP23:%.*]] = icmp ugt i32* [[TMP21]], [[TMP22]]
; THRESH-NEXT:    [[TMP24:%.*]] = select i1 [[TMP23]], i32* [[TMP21]], i32* [[TMP22]]
; THRESH-NEXT:    ret i32* [[TMP24]]
;
  %2 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 0), align 16
  %3 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 1), align 4
  %4 = icmp ugt i32* %2, %3
  %5 = select i1 %4, i32* %2, i32* %3
  %6 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 2), align 8
  %7 = icmp ugt i32* %5, %6
  %8 = select i1 %7, i32* %5, i32* %6
  %9 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 3), align 4
  %10 = icmp ugt i32* %8, %9
  %11 = select i1 %10, i32* %8, i32* %9
  %12 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 4), align 16
  %13 = icmp ugt i32* %11, %12
  %14 = select i1 %13, i32* %11, i32* %12
  %15 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 5), align 4
  %16 = icmp ugt i32* %14, %15
  %17 = select i1 %16, i32* %14, i32* %15
  %18 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 6), align 8
  %19 = icmp ugt i32* %17, %18
  %20 = select i1 %19, i32* %17, i32* %18
  %21 = load i32*, i32** getelementptr inbounds ([32 x i32*], [32 x i32*]* @arrp, i64 0, i64 7), align 4
  %22 = icmp ugt i32* %20, %21
  %23 = select i1 %22, i32* %20, i32* %21
  ret i32* %23
}

define i32 @smax_intrinsic_rdx_v8i32(i32* %p0) {
; SSE-LABEL: @smax_intrinsic_rdx_v8i32(
; SSE-NEXT:    [[P1:%.*]] = getelementptr inbounds i32, i32* [[P0:%.*]], i64 1
; SSE-NEXT:    [[P2:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 2
; SSE-NEXT:    [[P3:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 3
; SSE-NEXT:    [[P4:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 4
; SSE-NEXT:    [[P5:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 5
; SSE-NEXT:    [[P6:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 6
; SSE-NEXT:    [[P7:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 7
; SSE-NEXT:    [[T0:%.*]] = load i32, i32* [[P0]], align 4
; SSE-NEXT:    [[T1:%.*]] = load i32, i32* [[P1]], align 4
; SSE-NEXT:    [[T2:%.*]] = load i32, i32* [[P2]], align 4
; SSE-NEXT:    [[T3:%.*]] = load i32, i32* [[P3]], align 4
; SSE-NEXT:    [[T4:%.*]] = load i32, i32* [[P4]], align 4
; SSE-NEXT:    [[T5:%.*]] = load i32, i32* [[P5]], align 4
; SSE-NEXT:    [[T6:%.*]] = load i32, i32* [[P6]], align 4
; SSE-NEXT:    [[T7:%.*]] = load i32, i32* [[P7]], align 4
; SSE-NEXT:    [[M10:%.*]] = tail call i32 @llvm.smax.i32(i32 [[T1]], i32 [[T0]])
; SSE-NEXT:    [[M32:%.*]] = tail call i32 @llvm.smax.i32(i32 [[T3]], i32 [[T2]])
; SSE-NEXT:    [[M54:%.*]] = tail call i32 @llvm.smax.i32(i32 [[T5]], i32 [[T4]])
; SSE-NEXT:    [[M76:%.*]] = tail call i32 @llvm.smax.i32(i32 [[T7]], i32 [[T6]])
; SSE-NEXT:    [[M3210:%.*]] = tail call i32 @llvm.smax.i32(i32 [[M32]], i32 [[M10]])
; SSE-NEXT:    [[M7654:%.*]] = tail call i32 @llvm.smax.i32(i32 [[M76]], i32 [[M54]])
; SSE-NEXT:    [[M:%.*]] = tail call i32 @llvm.smax.i32(i32 [[M7654]], i32 [[M3210]])
; SSE-NEXT:    ret i32 [[M]]
;
; AVX-LABEL: @smax_intrinsic_rdx_v8i32(
; AVX-NEXT:    [[P1:%.*]] = getelementptr inbounds i32, i32* [[P0:%.*]], i64 1
; AVX-NEXT:    [[P2:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 2
; AVX-NEXT:    [[P3:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 3
; AVX-NEXT:    [[P4:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 4
; AVX-NEXT:    [[P5:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 5
; AVX-NEXT:    [[P6:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 6
; AVX-NEXT:    [[P7:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 7
; AVX-NEXT:    [[TMP1:%.*]] = bitcast i32* [[P0]] to <8 x i32>*
; AVX-NEXT:    [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* [[TMP1]], align 4
; AVX-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> [[TMP2]])
; AVX-NEXT:    ret i32 [[TMP3]]
;
; AVX2-LABEL: @smax_intrinsic_rdx_v8i32(
; AVX2-NEXT:    [[P1:%.*]] = getelementptr inbounds i32, i32* [[P0:%.*]], i64 1
; AVX2-NEXT:    [[P2:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 2
; AVX2-NEXT:    [[P3:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 3
; AVX2-NEXT:    [[P4:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 4
; AVX2-NEXT:    [[P5:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 5
; AVX2-NEXT:    [[P6:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 6
; AVX2-NEXT:    [[P7:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 7
; AVX2-NEXT:    [[TMP1:%.*]] = bitcast i32* [[P0]] to <8 x i32>*
; AVX2-NEXT:    [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* [[TMP1]], align 4
; AVX2-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> [[TMP2]])
; AVX2-NEXT:    ret i32 [[TMP3]]
;
; THRESH-LABEL: @smax_intrinsic_rdx_v8i32(
; THRESH-NEXT:    [[P1:%.*]] = getelementptr inbounds i32, i32* [[P0:%.*]], i64 1
; THRESH-NEXT:    [[P2:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 2
; THRESH-NEXT:    [[P3:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 3
; THRESH-NEXT:    [[P4:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 4
; THRESH-NEXT:    [[P5:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 5
; THRESH-NEXT:    [[P6:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 6
; THRESH-NEXT:    [[P7:%.*]] = getelementptr inbounds i32, i32* [[P0]], i64 7
; THRESH-NEXT:    [[TMP1:%.*]] = bitcast i32* [[P0]] to <8 x i32>*
; THRESH-NEXT:    [[TMP2:%.*]] = load <8 x i32>, <8 x i32>* [[TMP1]], align 4
; THRESH-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> [[TMP2]])
; THRESH-NEXT:    ret i32 [[TMP3]]
;
  %p1 = getelementptr inbounds i32, i32* %p0, i64 1
  %p2 = getelementptr inbounds i32, i32* %p0, i64 2
  %p3 = getelementptr inbounds i32, i32* %p0, i64 3
  %p4 = getelementptr inbounds i32, i32* %p0, i64 4
  %p5 = getelementptr inbounds i32, i32* %p0, i64 5
  %p6 = getelementptr inbounds i32, i32* %p0, i64 6
  %p7 = getelementptr inbounds i32, i32* %p0, i64 7
  %t0 = load i32, i32* %p0, align 4
  %t1 = load i32, i32* %p1, align 4
  %t2 = load i32, i32* %p2, align 4
  %t3 = load i32, i32* %p3, align 4
  %t4 = load i32, i32* %p4, align 4
  %t5 = load i32, i32* %p5, align 4
  %t6 = load i32, i32* %p6, align 4
  %t7 = load i32, i32* %p7, align 4
  %m10 = tail call i32 @llvm.smax.i32(i32 %t1, i32 %t0)
  %m32 = tail call i32 @llvm.smax.i32(i32 %t3, i32 %t2)
  %m54 = tail call i32 @llvm.smax.i32(i32 %t5, i32 %t4)
  %m76 = tail call i32 @llvm.smax.i32(i32 %t7, i32 %t6)
  %m3210 = tail call i32 @llvm.smax.i32(i32 %m32, i32 %m10)
  %m7654 = tail call i32 @llvm.smax.i32(i32 %m76, i32 %m54)
  %m = tail call i32 @llvm.smax.i32(i32 %m7654, i32 %m3210)
  ret i32 %m
}

define i16 @smin_intrinsic_rdx_v8i16(i16* %p0) {
; CHECK-LABEL: @smin_intrinsic_rdx_v8i16(
; CHECK-NEXT:    [[P1:%.*]] = getelementptr inbounds i16, i16* [[P0:%.*]], i64 1
; CHECK-NEXT:    [[P2:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 2
; CHECK-NEXT:    [[P3:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 3
; CHECK-NEXT:    [[P4:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 4
; CHECK-NEXT:    [[P5:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 5
; CHECK-NEXT:    [[P6:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 6
; CHECK-NEXT:    [[P7:%.*]] = getelementptr inbounds i16, i16* [[P0]], i64 7
; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i16* [[P0]] to <8 x i16>*
; CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i16>, <8 x i16>* [[TMP1]], align 4
; CHECK-NEXT:    [[TMP3:%.*]] = call i16 @llvm.vector.reduce.smin.v8i16(<8 x i16> [[TMP2]])
; CHECK-NEXT:    ret i16 [[TMP3]]
;
  %p1 = getelementptr inbounds i16, i16* %p0, i64 1
  %p2 = getelementptr inbounds i16, i16* %p0, i64 2
  %p3 = getelementptr inbounds i16, i16* %p0, i64 3
  %p4 = getelementptr inbounds i16, i16* %p0, i64 4
  %p5 = getelementptr inbounds i16, i16* %p0, i64 5
  %p6 = getelementptr inbounds i16, i16* %p0, i64 6
  %p7 = getelementptr inbounds i16, i16* %p0, i64 7
  %t0 = load i16, i16* %p0, align 4
  %t1 = load i16, i16* %p1, align 4
  %t2 = load i16, i16* %p2, align 4
  %t3 = load i16, i16* %p3, align 4
  %t4 = load i16, i16* %p4, align 4
  %t5 = load i16, i16* %p5, align 4
  %t6 = load i16, i16* %p6, align 4
  %t7 = load i16, i16* %p7, align 4
  %m10 = tail call i16 @llvm.smin.i16(i16 %t1, i16 %t0)
  %m32 = tail call i16 @llvm.smin.i16(i16 %t3, i16 %t2)
  %m54 = tail call i16 @llvm.smin.i16(i16 %t5, i16 %t4)
  %m76 = tail call i16 @llvm.smin.i16(i16 %t7, i16 %t6)
  %m3210 = tail call i16 @llvm.smin.i16(i16 %m32, i16 %m10)
  %m7654 = tail call i16 @llvm.smin.i16(i16 %m76, i16 %m54)
  %m = tail call i16 @llvm.smin.i16(i16 %m7654, i16 %m3210)
  ret i16 %m
}

define i64 @umax_intrinsic_rdx_v4i64(i64* %p0) {
; DEFAULT-LABEL: @umax_intrinsic_rdx_v4i64(
; DEFAULT-NEXT:    [[P1:%.*]] = getelementptr inbounds i64, i64* [[P0:%.*]], i64 1
; DEFAULT-NEXT:    [[P2:%.*]] = getelementptr inbounds i64, i64* [[P0]], i64 2
; DEFAULT-NEXT:    [[P3:%.*]] = getelementptr inbounds i64, i64* [[P0]], i64 3
; DEFAULT-NEXT:    [[T0:%.*]] = load i64, i64* [[P0]], align 4
; DEFAULT-NEXT:    [[T1:%.*]] = load i64, i64* [[P1]], align 4
; DEFAULT-NEXT:    [[T2:%.*]] = load i64, i64* [[P2]], align 4
; DEFAULT-NEXT:    [[T3:%.*]] = load i64, i64* [[P3]], align 4
; DEFAULT-NEXT:    [[M10:%.*]] = tail call i64 @llvm.umax.i64(i64 [[T1]], i64 [[T0]])
; DEFAULT-NEXT:    [[M32:%.*]] = tail call i64 @llvm.umax.i64(i64 [[T3]], i64 [[T2]])
; DEFAULT-NEXT:    [[M:%.*]] = tail call i64 @llvm.umax.i64(i64 [[M32]], i64 [[M10]])
; DEFAULT-NEXT:    ret i64 [[M]]
;
; THRESH-LABEL: @umax_intrinsic_rdx_v4i64(
; THRESH-NEXT:    [[P1:%.*]] = getelementptr inbounds i64, i64* [[P0:%.*]], i64 1
; THRESH-NEXT:    [[P2:%.*]] = getelementptr inbounds i64, i64* [[P0]], i64 2
; THRESH-NEXT:    [[P3:%.*]] = getelementptr inbounds i64, i64* [[P0]], i64 3
; THRESH-NEXT:    [[TMP1:%.*]] = bitcast i64* [[P0]] to <4 x i64>*
; THRESH-NEXT:    [[TMP2:%.*]] = load <4 x i64>, <4 x i64>* [[TMP1]], align 4
; THRESH-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vector.reduce.umax.v4i64(<4 x i64> [[TMP2]])
; THRESH-NEXT:    ret i64 [[TMP3]]
;
  %p1 = getelementptr inbounds i64, i64* %p0, i64 1
  %p2 = getelementptr inbounds i64, i64* %p0, i64 2
  %p3 = getelementptr inbounds i64, i64* %p0, i64 3
  %t0 = load i64, i64* %p0, align 4
  %t1 = load i64, i64* %p1, align 4
  %t2 = load i64, i64* %p2, align 4
  %t3 = load i64, i64* %p3, align 4
  %m10 = tail call i64 @llvm.umax.i64(i64 %t1, i64 %t0)
  %m32 = tail call i64 @llvm.umax.i64(i64 %t3, i64 %t2)
  %m = tail call i64 @llvm.umax.i64(i64 %m32, i64 %m10)
  ret i64 %m
}

define i8 @umin_intrinsic_rdx_v16i8(i8* %p0) {
; CHECK-LABEL: @umin_intrinsic_rdx_v16i8(
; CHECK-NEXT:    [[P1:%.*]] = getelementptr inbounds i8, i8* [[P0:%.*]], i64 1
; CHECK-NEXT:    [[P2:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 2
; CHECK-NEXT:    [[P3:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 3
; CHECK-NEXT:    [[P4:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 4
; CHECK-NEXT:    [[P5:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 5
; CHECK-NEXT:    [[P6:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 6
; CHECK-NEXT:    [[P7:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 7
; CHECK-NEXT:    [[P8:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 8
; CHECK-NEXT:    [[P9:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 9
; CHECK-NEXT:    [[PA:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 10
; CHECK-NEXT:    [[PB:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 11
; CHECK-NEXT:    [[PC:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 12
; CHECK-NEXT:    [[PD:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 13
; CHECK-NEXT:    [[PE:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 14
; CHECK-NEXT:    [[PF:%.*]] = getelementptr inbounds i8, i8* [[P0]], i64 15
; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[P0]] to <16 x i8>*
; CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[TMP1]], align 4
; CHECK-NEXT:    [[TMP3:%.*]] = call i8 @llvm.vector.reduce.umin.v16i8(<16 x i8> [[TMP2]])
; CHECK-NEXT:    ret i8 [[TMP3]]
;
  %p1 = getelementptr inbounds i8, i8* %p0, i64 1
  %p2 = getelementptr inbounds i8, i8* %p0, i64 2
  %p3 = getelementptr inbounds i8, i8* %p0, i64 3
  %p4 = getelementptr inbounds i8, i8* %p0, i64 4
  %p5 = getelementptr inbounds i8, i8* %p0, i64 5
  %p6 = getelementptr inbounds i8, i8* %p0, i64 6
  %p7 = getelementptr inbounds i8, i8* %p0, i64 7
  %p8 = getelementptr inbounds i8, i8* %p0, i64 8
  %p9 = getelementptr inbounds i8, i8* %p0, i64 9
  %pa = getelementptr inbounds i8, i8* %p0, i64 10
  %pb = getelementptr inbounds i8, i8* %p0, i64 11
  %pc = getelementptr inbounds i8, i8* %p0, i64 12
  %pd = getelementptr inbounds i8, i8* %p0, i64 13
  %pe = getelementptr inbounds i8, i8* %p0, i64 14
  %pf = getelementptr inbounds i8, i8* %p0, i64 15
  %t0 = load i8, i8* %p0, align 4
  %t1 = load i8, i8* %p1, align 4
  %t2 = load i8, i8* %p2, align 4
  %t3 = load i8, i8* %p3, align 4
  %t4 = load i8, i8* %p4, align 4
  %t5 = load i8, i8* %p5, align 4
  %t6 = load i8, i8* %p6, align 4
  %t7 = load i8, i8* %p7, align 4
  %t8 = load i8, i8* %p8, align 4
  %t9 = load i8, i8* %p9, align 4
  %ta = load i8, i8* %pa, align 4
  %tb = load i8, i8* %pb, align 4
  %tc = load i8, i8* %pc, align 4
  %td = load i8, i8* %pd, align 4
  %te = load i8, i8* %pe, align 4
  %tf = load i8, i8* %pf, align 4
  %m10 = tail call i8 @llvm.umin.i8(i8 %t1, i8 %t0)
  %m32 = tail call i8 @llvm.umin.i8(i8 %t3, i8 %t2)
  %m54 = tail call i8 @llvm.umin.i8(i8 %t5, i8 %t4)
  %m76 = tail call i8 @llvm.umin.i8(i8 %t7, i8 %t6)
  %m98 = tail call i8 @llvm.umin.i8(i8 %t9, i8 %t8)
  %mba = tail call i8 @llvm.umin.i8(i8 %tb, i8 %ta)
  %mdc = tail call i8 @llvm.umin.i8(i8 %td, i8 %tc)
  %mfe = tail call i8 @llvm.umin.i8(i8 %tf, i8 %te)
  %m3210 = tail call i8 @llvm.umin.i8(i8 %m32, i8 %m10)
  %m7654 = tail call i8 @llvm.umin.i8(i8 %m76, i8 %m54)
  %mdc98 = tail call i8 @llvm.umin.i8(i8 %mdc, i8 %m98)
  %mfeba = tail call i8 @llvm.umin.i8(i8 %mfe, i8 %mba)
  %ml = tail call i8 @llvm.umin.i8(i8 %m3210, i8 %m7654)
  %mh = tail call i8 @llvm.umin.i8(i8 %mfeba, i8 %mdc98)
  %m = tail call i8 @llvm.umin.i8(i8 %mh, i8 %ml)
  ret i8 %m
}

; This should not crash.

define void @PR49730() {
; SSE-LABEL: @PR49730(
; SSE-NEXT:    [[T:%.*]] = call i32 @llvm.smin.i32(i32 undef, i32 2)
; SSE-NEXT:    [[T1:%.*]] = sub nsw i32 undef, [[T]]
; SSE-NEXT:    [[T2:%.*]] = call i32 @llvm.umin.i32(i32 undef, i32 [[T1]])
; SSE-NEXT:    [[T3:%.*]] = call i32 @llvm.smin.i32(i32 undef, i32 2)
; SSE-NEXT:    [[T4:%.*]] = sub nsw i32 undef, [[T3]]
; SSE-NEXT:    [[T5:%.*]] = call i32 @llvm.umin.i32(i32 [[T2]], i32 [[T4]])
; SSE-NEXT:    [[T6:%.*]] = call i32 @llvm.smin.i32(i32 undef, i32 1)
; SSE-NEXT:    [[T7:%.*]] = sub nuw nsw i32 undef, [[T6]]
; SSE-NEXT:    [[T8:%.*]] = call i32 @llvm.umin.i32(i32 [[T5]], i32 [[T7]])
; SSE-NEXT:    [[T9:%.*]] = call i32 @llvm.smin.i32(i32 undef, i32 1)
; SSE-NEXT:    [[T10:%.*]] = sub nsw i32 undef, [[T9]]
; SSE-NEXT:    [[T11:%.*]] = call i32 @llvm.umin.i32(i32 [[T8]], i32 [[T10]])
; SSE-NEXT:    [[T12:%.*]] = sub nsw i32 undef, undef
; SSE-NEXT:    [[T13:%.*]] = call i32 @llvm.umin.i32(i32 [[T11]], i32 [[T12]])
; SSE-NEXT:    [[T14:%.*]] = call i32 @llvm.umin.i32(i32 [[T13]], i32 93)
; SSE-NEXT:    ret void
;
; AVX-LABEL: @PR49730(
; AVX-NEXT:    [[TMP1:%.*]] = call <4 x i32> @llvm.smin.v4i32(<4 x i32> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>)
; AVX-NEXT:    [[TMP2:%.*]] = sub nsw <4 x i32> poison, [[TMP1]]
; AVX-NEXT:    [[T12:%.*]] = sub nsw i32 undef, undef
; AVX-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.umin.v4i32(<4 x i32> [[TMP2]])
; AVX-NEXT:    [[TMP4:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP3]], i32 [[T12]])
; AVX-NEXT:    [[TMP5:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP4]], i32 undef)
; AVX-NEXT:    [[T14:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP5]], i32 93)
; AVX-NEXT:    ret void
;
; AVX2-LABEL: @PR49730(
; AVX2-NEXT:    [[TMP1:%.*]] = call <4 x i32> @llvm.smin.v4i32(<4 x i32> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>)
; AVX2-NEXT:    [[TMP2:%.*]] = sub nsw <4 x i32> poison, [[TMP1]]
; AVX2-NEXT:    [[T12:%.*]] = sub nsw i32 undef, undef
; AVX2-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.umin.v4i32(<4 x i32> [[TMP2]])
; AVX2-NEXT:    [[TMP4:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP3]], i32 [[T12]])
; AVX2-NEXT:    [[TMP5:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP4]], i32 undef)
; AVX2-NEXT:    [[T14:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP5]], i32 93)
; AVX2-NEXT:    ret void
;
; THRESH-LABEL: @PR49730(
; THRESH-NEXT:    [[TMP1:%.*]] = call <4 x i32> @llvm.smin.v4i32(<4 x i32> poison, <4 x i32> <i32 2, i32 2, i32 1, i32 1>)
; THRESH-NEXT:    [[TMP2:%.*]] = sub nsw <4 x i32> poison, [[TMP1]]
; THRESH-NEXT:    [[T12:%.*]] = sub nsw i32 undef, undef
; THRESH-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.umin.v4i32(<4 x i32> [[TMP2]])
; THRESH-NEXT:    [[TMP4:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP3]], i32 [[T12]])
; THRESH-NEXT:    [[TMP5:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP4]], i32 undef)
; THRESH-NEXT:    [[T14:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP5]], i32 93)
; THRESH-NEXT:    ret void
;
  %t = call i32 @llvm.smin.i32(i32 undef, i32 2)
  %t1 = sub nsw i32 undef, %t
  %t2 = call i32 @llvm.umin.i32(i32 undef, i32 %t1)
  %t3 = call i32 @llvm.smin.i32(i32 undef, i32 2)
  %t4 = sub nsw i32 undef, %t3
  %t5 = call i32 @llvm.umin.i32(i32 %t2, i32 %t4)
  %t6 = call i32 @llvm.smin.i32(i32 undef, i32 1)
  %t7 = sub nuw nsw i32 undef, %t6
  %t8 = call i32 @llvm.umin.i32(i32 %t5, i32 %t7)
  %t9 = call i32 @llvm.smin.i32(i32 undef, i32 1)
  %t10 = sub nsw i32 undef, %t9
  %t11 = call i32 @llvm.umin.i32(i32 %t8, i32 %t10)
  %t12 = sub nsw i32 undef, undef
  %t13 = call i32 @llvm.umin.i32(i32 %t11, i32 %t12)
  %t14 = call i32 @llvm.umin.i32(i32 %t13, i32 93)
  ret void
}